Every major AI API price. One page. Then what the same task actually costs.
List prices for OpenAI, Anthropic, Google, xAI and DeepSeek, pulled from the vendors’ own pricing pages and checked July 7, 2026. Then the part those pages leave out: the same job, nine models, and what each passing grade actually cost — so you can see where the savings come from.
The list prices
Every major API price. One page.
USD per 1M tokens, standard tiers, as published by each vendor. Checked July 7, 2026.
| Model | $ in / Mtok | $ out / Mtok |
|---|
One-line source note: list prices copied from each vendor’s public pricing page — openai.com, claude.com (Anthropic), ai.google.dev, x.ai, deepseek.com · checked July 7, 2026.
Fable 5 left Claude subscriptions on July 7, 2026 — it is API-only now.
The catch
List price isn’t cost. Token burn is.
A pricing page tells you the rate. Your bill is rate × tokens burned to a passing grade — so we ran the same task on nine models, one shot each, and published the meter.
Five of nine models passed the same 8-point audit. 39 times the price spread for the same passing grade.
Sometimes the pricey one is worth it: GPT 5.5 shipped the best-designed page of the run. That’s a KEEP — and a KEEP pays us $0.
Run №000 · 07.02.26 · 8-point Playwright audit, one shot each, nine models. Four DNF’d on transport errors — Fable 5, Opus 4.8, Kimi K2.5, MiniMax M2.7 — failures in the pipes, not the models. №000 is our open run; the blind ledger holds contests №001–005 ↗
Receipts
Same prompt. Both answers. You judge.
Pick a prompt. Two answers come back — one from the frontier model, one from the value model, with the real cost on each. You decide whether the grade holds.
Real runs through our own gateway, July 7 2026 · answers untouched · costs are what the meter read
The mechanism
Proof, not promises.
Point your OpenAI and Anthropic keys at our pass-through proxy. Same models, same outputs, read-only. You keep your provider — nothing to rip out.
It re-runs a sample of your real requests async, off the production path. Users never see it.
Quality and price per workload, re-gradable from the transcripts. Nothing moves without your sign-off.
Pareto wasn’t ready to enter Run №000 — once it ships, it enters every contest under the same audit.
Live July 15
Send us your AI bill.
We’ll show you the difference.
We’ll write when unbiased.ai opens on July 15. First shadow runs go to the list in order.
Join before July 15 → $50 in credits when we open · keys encrypted, never logged · free until you flip Pareto to production
Prefer email? Send your bill to hello@unbiased.ai